no applicable method for 'smoothType' applied to an object of class "c('adam', 'smooth')"

[mark0100]

When I run the example code from smoothCombine() i get the following error:

models <- list(es(BJsales), sma(BJsales))
smoothCombine(BJsales,models=models)
Error in UseMethod("smoothType") :
no applicable method for 'smoothType' applied to an object of class "c('adam', 'smooth')"

traceback()
4: FUN(X[[i]], ...)
3: lapply(models, smoothType)
2: unlist(lapply(models, smoothType))
1: smoothCombine(BJsales, models = models)

I also get this error with my own data when i call smoothCombine()
I tried detaching/removing and reinstalling the smooth package from CRAN but still the error persists.

I am on linux with R 4.3 and:

library(smooth)
This is package "smooth", v3.2.0

If i look in the code for smoothCombine() there is a reference to a variable "smoothType" (on line 99) but it is never set:

if (is.null(names(models))) {
    names(models) <- unlist(lapply(models, smoothType))
}

And i can confirm that there are no names in the models object:

models <- list(es(BJsales), sma(BJsales))
names(models)
NULL

So if i run this code snippet manually i get the same error:

unlist(lapply(models, smoothType))
Error: object 'smoothType' not found

What should smoothType be? Do i have to set it manually before calling smoothCombine()?

Making the models named parameters does solve the smoothType error but throws another error:

models <- list(es=es(BJsales), sma=sma(BJsales))
smoothCombine(BJsales,models=models)
Error in match.arg(interval[1], c("none", "simulated", "approximate", :
'arg' must be NULL or a character vector

traceback()
6: stop("'arg' must be NULL or a character vector")
5: match.arg(interval[1], c("none", "simulated", "approximate",
"semiparametric", "nonparametric", "confidence", "parametric",
"prediction", "empirical", "complete"))
4: forecast.adam(X[[i]], ...)
3: FUN(X[[i]], ...)
2: lapply(models, forecast, h = h, interval = interval, level = 0,
holdout = holdout, cumulative = cumulative, xreg = xreg)
1: smoothCombine(BJsales, models = models)

[config-i1]

Hi,

Thanks for your message! smoothType is an internal function that extracts the type of model. It looks like a bug in how it is defined and how it is exported from the package. I'll fix this.

Also, I noticed an issue with the interval type. This is now fixed in d49850c If you want to have the fix, you'll need to install the package from github.

[config-i1]

Should be fixed now in db1546e Please check and let me know if the problem (or a similar one) persists.

[mark0100]

Hello Ivan,

Thanks for your quick reply.

I have tested the new version and the standard example for smoothCombine() now works.
The smoothType error is fixed but with some data and parameter settings i still get the match.arg error on interval when calling smoothCombine():

models <- list(es(BJsales, h=12, holdout=TRUE, interval="p", silent=FALSE), sma(BJsales, h=12, holdout=TRUE, interval="p", silent=FALSE))
Forming the pool of models based on... ANN, AAN, Estimation progress: 100%... Done!
Order 1 - 510.772; Order 69 - 1173.5638; Order 138 - 1250.2688
Order 1 - 510.772; Order 35 - 1088.5239; Order 69 - 1173.5638
Order 1 - 510.772; Order 18 - 965.1295; Order 35 - 1088.5239
Order 1 - 510.772; Order 9 - 831.8731; Order 18 - 965.1295
Order 1 - 510.772; Order 5 - 719.3918; Order 9 - 831.8731
Order 1 - 510.772; Order 3 - 632.8999; Order 5 - 719.3918
Order 1 - 510.772; Order 2 - 576.5735; Order 3 - 632.8999
Warning message:
In es(BJsales, h = 12, holdout = TRUE, interval = "p", silent = FALSE) :
Parameter "interval" is no longer supported in es(). Please use forecast() method to produce prediction interval.

smoothCombine(BJsales,models=models, h=12, holdout=TRUE, interval="p", silent=FALSE)
Error in match.arg(interval[1], c("none", "simulated", "approximate", :
'arg' should be one of “none”, “simulated”, “approximate”, “semiparametric”, “nonparametric”, “confidence”, “parametric”, “prediction”, “empirical”, “complete”

traceback()
6: stop(sprintf(ngettext(length(chs <- unique(choices[nzchar(choices)])),
"'arg' should be %s", "'arg' should be one of %s"), paste(dQuote(chs),
collapse = ", ")), domain = NA)
5: match.arg(interval[1], c("none", "simulated", "approximate",
"semiparametric", "nonparametric", "confidence", "parametric",
"prediction", "empirical", "complete"))
4: forecast.adam(X[[i]], ...)
3: FUN(X[[i]], ...)
2: lapply(models, forecast, h = h, interval = intervalOriginal,
level = 0, holdout = holdout, cumulative = cumulative, xreg = xreg)
1: smoothCombine(BJsales, models = models, h = 12, holdout = TRUE,
interval = "p", silent = FALSE)

I am not sure this is still a bug or just me providing wrong parameters/data.

[config-i1]

Can you please try interval="parametric" instead? There have been some changes in es() and sma() functions. As you see it no longer accepts interval as one of parameters. I haven't updated the smoothCombine() properly to align it with the new functions, but it should not matter much.

[mark0100]

I have tried that but it makes no difference:

models <- list(es(BJsales, h=12, holdout=TRUE, interval="parametric", silent=FALSE), sma(BJsales, h=12, holdout=TRUE, interval="parametric", silent=FALSE))
Forming the pool of models based on... ANN, AAN, Estimation progress: 100%... Done!
Order 1 - 510.772; Order 69 - 1173.5638; Order 138 - 1250.2688
Order 1 - 510.772; Order 35 - 1088.5239; Order 69 - 1173.5638
Order 1 - 510.772; Order 18 - 965.1295; Order 35 - 1088.5239
Order 1 - 510.772; Order 9 - 831.8731; Order 18 - 965.1295
Order 1 - 510.772; Order 5 - 719.3918; Order 9 - 831.8731
Order 1 - 510.772; Order 3 - 632.8999; Order 5 - 719.3918
Order 1 - 510.772; Order 2 - 576.5735; Order 3 - 632.8999
Warning message:
In es(BJsales, h = 12, holdout = TRUE, interval = "parametric", :
Parameter "interval" is no longer supported in es(). Please use forecast() method to produce prediction interval.

smoothCombine(BJsales,models=models, h=12, holdout=TRUE, interval="parametric", silent=FALSE)
Error in match.arg(interval[1], c("none", "simulated", "approximate", :
'arg' must be NULL or a character vector

[config-i1]

Thanks! I'll have a look when I have free time and try to fix it. Apologies for inconvenience.